Centrifugal chiller surge: causes, control, and how to stop it

What surge is, and why it matters

Centrifugal chiller surge is a momentary reversal of refrigerant flow through the compressor, and it sounds like the machine is trying to come apart. A centrifugal compressor is a dynamic machine. It does not trap and squeeze a fixed pocket of gas the way a screw or scroll does. It spins an impeller that flings refrigerant outward and converts that velocity into pressure, and the pressure it can build at a given speed and flow has a ceiling. When the lift the machine has to make, the gap between evaporator pressure and condenser pressure, climbs past what the impeller can produce at the flow it is moving, the gas that was just pushed forward stalls, turns around, and slams back through the wheel.

Then the compressor recovers, builds head again, and the whole event repeats a second or two later. That repeating reversal is surge. You hear it as a deep bang or a rolling rumble, you feel it in the floor, and you watch the motor amps and the discharge temperature jump with each cycle.

It matters because a centrifugal machine is not built to take flow going the wrong way. Each reversal slams the rotor against its thrust bearing and shakes the impeller. A few surges on a hard start are survivable. A machine left to ride through surge for an afternoon is a machine with a wiped thrust bearing and, eventually, a cracked or rubbed impeller. Of the common compressor families covered in the chiller-types guide, the centrifugal is the one that can surge, which is the price of being the efficiency leader at large tonnage. Where surge shows up first is during startup and at low load, which is why the startup-and-commissioning guide treats it as a gate to clear before acceptance.

Why a centrifugal compressor surges: head, not displacement

The physics starts with what kind of compressor this is. A positive-displacement machine, a screw or a scroll, captures a volume of gas and mechanically reduces it, so it will build whatever pressure it is asked to build until something breaks or the motor stalls. A centrifugal does not work that way. It adds kinetic energy to the gas with a spinning impeller and then trades that velocity for pressure in the diffuser. The head it can develop depends on tip speed and on how much gas is flowing through the wheel.

Picture the operating point as a balance. The machine produces a certain head at a certain flow, and the system, the evaporator on one side and the condenser on the other, demands a certain lift. As long as the impeller can make more head than the system asks for, flow stays forward and stable. Push the demanded lift up, or let the flow fall, and you walk toward the point where the impeller can just barely make the head. One step past it, the impeller cannot hold the discharge pressure back.

When that happens the high-pressure gas in the condenser shoves backward through the wheel. Flow reverses for a fraction of a second, the discharge pressure bleeds off, the impeller regains control and pushes flow forward again, pressure rebuilds, and the gas reverses a second time. The reversal is not a one-time event. It is an oscillation that will continue as long as the machine is asked to make more lift than it can at the flow it has. That distinction, head versus displacement, is the whole reason a centrifugal can surge and a screw cannot.

The compressor map and the surge line

Every centrifugal compressor has a map, and the manufacturer builds the surge control around it. The map plots head, the lift in feet or in pressure terms, against flow, with a family of curves for different speeds or vane positions. Two limits box in the usable area. On the left is the surge line, the locus of low-flow, high-head points where the machine becomes unstable. On the right is choke, sometimes called stonewall, where gas reaches sonic velocity in the passages and flow simply cannot increase no matter how much you drop the downstream pressure.

Operation lives in the envelope between those two lines. Most chiller operation never goes near choke. Surge is the limit that bites, because the two things that move the operating point left, toward the surge line, are exactly the two things that happen on a real plant: the lift rises, or the load and flow fall. Cross the line and you are in surge.

The map is the manufacturer's property and it is specific to the machine, the refrigerant, and the impeller trim. Do not carry a rule from one chiller to another. The point of knowing the map exists is to understand what the control is doing when it opens the vanes or slows the compressor: it is dragging the operating point back to the right of the surge line.

How do you know a chiller is surging?

The bang is the tell. A surging centrifugal makes a loud, low banging or a rolling rumble, sometimes described as a whoosh and reverse, and it comes in cycles a second or two apart rather than once. Stand in the machine room during a surge and you will not mistake it for anything else. If you can hear it from outside the room, it has been going long enough to do damage.

The instruments move with the sound. Motor amps swing up and down in step with each reversal, often a wide swing rather than a flutter, because the load on the impeller is collapsing and rebuilding. Discharge or condenser pressure and the discharge temperature jump with the same rhythm. On a machine with good logging you can see the surge in the amp trace before anyone reports the noise.

Vibration rises, and that is the part that costs money. The rotor is being slammed axially against the thrust bearing on each cycle, so a surging machine shows up later as thrust-bearing wear, babbitt damage, or measurable end-play that was not there at the last oil analysis. A machine that surged briefly on startup and then settled is one thing. A machine that surges every time it stages down to low load is telling you it has a control or a lift problem that is grinding the bearing a little more each day.

The real cause: high lift

High lift causes surge. That is the sentence to keep at the front of the diagnosis. Lift is the pressure difference the compressor has to bridge, set by the condensing pressure on top and the evaporator pressure on the bottom. Drive either end the wrong way and the machine has to make more head, and at any given flow it is closer to the surge line.

In the field the condenser side is almost always the culprit, because that is where lift gets added by problems that creep in slowly. High condensing pressure comes from dirty condenser tubes that will not transfer heat, from non-condensable gas blanketing the tubes, from condenser water that comes back from the tower too warm, or from low condenser-water flow that lets the water heat up too much as it passes through. Each one raises the condensing temperature and pressure, and every bit of that raise is lift the compressor now has to produce.

The evaporator side adds lift too when the chilled-water setpoint is pushed low or the evaporator is fouled, because a colder, lower suction pressure widens the same gap. Add a low chilled-water setpoint to a warm condenser on a hot afternoon and you have manufactured a high-lift condition that walks the machine toward surge at part load. When a chiller that ran clean for years starts surging, the question is not usually the compressor. It is what raised the lift.

Low load and low flow push toward the surge line

The other way to reach the surge line is from the flow side. As the building load drops, the chiller has to move less refrigerant, so the control reduces flow through the impeller. Reduce flow far enough and the operating point slides left across the map until it touches the surge line, even when the lift is reasonable. This is why surge is a low-load phenomenon as much as a high-lift one. The two work together.

A light building is the common version. On a mild day, or overnight, the load can fall to a fraction of design while a single large machine is still the only one running. The chiller throttles down, flow falls, and a fixed-speed machine with only inlet vanes for capacity control can run out of stable turndown before it runs out of building. Low evaporator-water flow does the same thing from the other direction, starving the machine of the flow it needs to stay off the line.

The fix for the flow side is not the same as the fix for the lift side, which is why the diagnosis has to name which one you have. Low load wants either a way to slow the compressor, a way to fake a load, or a smaller machine staged in its place. High lift wants the head brought back down. Confuse them and you spend money on the wrong cure.

The low condenser-water paradox

Operators sometimes blame cold tower water for surge, and it is worth being clear: low condenser-water temperature does not cause surge. It does the opposite. Colder condenser water lowers the condensing pressure, which lowers the lift, which gives the machine more room above the surge line, not less. It is also the single biggest efficiency lever on a water-cooled chiller, which is why condenser-water reset and a well-run tower save so much energy.

Where the confusion comes from is the low-load corner. A machine running at very low load with very low lift can still surge, because the problem there is flow, not head. The control may struggle to hold a stable point when both load and lift are low at once, and an operator watching cold tower water at the moment of a surge can pin the blame on the wrong variable. The cold water did not cause it. The collapse of flow at low load did, and the cold water was the thing keeping the rest of the envelope healthy.

So push condenser water as low as the chiller and the tower will allow for efficiency, within the minimum entering-condenser-water temperature the manufacturer sets. Just understand that at the bottom of the load range you may still need a VFD or hot-gas bypass to hold flow. The lever for surge is lift and flow, and high lift is the one that drives it. Do not back off good condenser-water reset chasing a surge that came from somewhere else.

What are inlet guide vanes?

Inlet guide vanes are a ring of adjustable vanes at the impeller eye that modulate the chiller's capacity by pre-swirling the refrigerant as it enters the wheel. Spinning the gas in the direction the impeller turns reduces how much work the impeller does on it, which reduces capacity smoothly without changing the motor speed. On a fixed-speed centrifugal, the IGV is the primary capacity control, and the position of those vanes is what moves the operating point around the map.

The catch is that closing the vanes to unload the machine throttles the flow down, and throttling flow down is exactly the move that walks the operating point toward the surge line. So the IGV both controls capacity and, if it goes too far at high lift, triggers the surge it is supposed to avoid. The control logic has to coordinate vane position against the available surge margin, not just against the leaving chilled-water temperature.

When a machine surges at part load, the IGV linkage and calibration are on the short list to check. A vane actuator that is out of calibration, a linkage that has slipped, or a feedback signal that does not match true vane position will let the control close the vanes further than the map allows. The vanes read one position on the panel and sit at another in the machine, and the surge that follows looks like a mystery until someone confirms the actual vane angle against the command.

The VFD chiller and surge margin

A variable-frequency drive changes the picture by letting the compressor slow down. At low lift the machine does not need full impeller tip speed to make the head it requires, so the drive reduces speed, which cuts energy use sharply and keeps the operating point off the surge line at part load. The affinity relationship between speed and power is why a VFD centrifugal at low lift and part load can post the lowest kW per ton in the plant.

The strongest control pairs the VFD with the inlet guide vanes. The drive handles the broad capacity change by speed, and the vanes trim and hold surge margin at the edges, especially at low load where speed alone runs out of room. Together they keep the point inside the envelope across a wide range of load and lift that a vane-only machine could not cover without surging. The chiller-types guide goes deeper on why this combination wins at scale.

A VFD is not a license to ignore lift, though. If the condenser is fouled and the lift is high, the drive has to run the compressor faster to make the head, the part-load savings evaporate, and you can still surge if the load also drops. The drive widens the stable envelope. It does not raise the condensing pressure ceiling that high lift creates. Clean tubes still come first.

Hot-gas bypass at very low load

Hot-gas bypass keeps a centrifugal off the surge line at very low load by faking a load on the machine. A valve routes hot discharge gas from the compressor outlet back to the evaporator or suction side, which adds an artificial heat load the compressor has to handle. That keeps the flow through the impeller up and holds the operating point to the right of the surge line even when the real building load has fallen below the machine's stable turndown.

It works, and on a fixed-speed machine that has to hold a building at very light load it is sometimes the only practical way to keep a single large chiller running without surging. It is the control of last resort at the bottom of the range, the floor under the turndown.

It also burns energy on purpose, and that is the honest tradeoff. Every bit of hot gas you bypass is compressor work spent moving heat in a circle instead of cooling the building, so a chiller leaning on hot-gas bypass is running at poor efficiency by design. The better answers, where the plant allows them, are a VFD for turndown or staging a smaller machine to match the light load, and using hot-gas bypass only for the sliver of operation below what those can cover. A plant that runs on bypass for hours every night was sized or staged wrong.

How surge protection control works

The chiller's control panel watches for surge and acts before the machine destroys itself. The exact method varies by manufacturer, but the logic is consistent: the panel monitors the operating point or the signature of a surge, opens the inlet guide vanes or raises speed to push flow back up and lift down, and if the surges keep coming it stages the machine down or trips it on a surge fault.

Many controls count surges. A single surge on a hard start may be logged and tolerated, but a defined number of surges within a window will force the machine offline rather than let it grind. That count is a protection feature, and resetting and restarting a machine that keeps tripping on surge without finding the cause is how thrust bearings die. The trip is the machine telling you the operating point is wrong, not a nuisance to clear.

Surge detection is only as good as its inputs. The control leans on pressure and temperature sensors and on the vane or speed feedback, and a drifted sensor or a miscalibrated vane feedback can either hide a real surge or trip on a phantom one. When a machine starts surging or starts false-tripping, verify the sensors and the feedback before you touch the surge-control setpoints. Changing protection setpoints to silence an alarm, instead of fixing what moved the operating point, is the most expensive mistake on this whole list.

Non-condensables and the purge unit

Non-condensable gas, mostly air and the moisture that comes with it, raises condensing pressure and is a leading hidden cause of surge. Air collects in the condenser and blankets the tubes, taking up space the refrigerant needs and degrading heat transfer, so the head the compressor must make climbs even though nothing mechanical has changed. The lift goes up, the surge margin goes down, and the cause is invisible until you look for it.

Low-pressure machines are the ones that pull air in. Chillers using low-pressure refrigerants such as R-123 and its replacement R-1233zd run the evaporator below atmospheric pressure, so any leak draws air inward rather than leaking refrigerant out. Those machines carry a purge unit whose job is to pull the accumulated non-condensables out of the condenser and vent them, keeping the head down. A purge that runs constantly is a leak you have not found yet, and the purge run-time log is one of the most useful early warnings of a developing problem.

Positive-pressure machines, such as those on R-134a, R-513A, and similar refrigerants, sit above atmospheric on the low side and so leak refrigerant out rather than drawing air in, and most do not carry a purge. There the non-condensable question shows up after the system has been opened for service and not properly evacuated. Either way, air in the condenser reads as rising lift, and rising lift reads as shrinking surge margin.

Condenser tube fouling and the approach diagnostic

Dirty condenser tubes raise the head, and a rising approach temperature is how you catch it before it shows up as surge. Scale, biofilm, mud, and silt from the tower water build a layer on the inside of the condenser tubes that the heat has to pass through, so the refrigerant condenses at a higher temperature and pressure than it should for the water temperature it is rejecting to. That added condensing pressure is added lift, and added lift is lost surge margin.

The approach temperature is the diagnostic that makes fouling visible. Condenser approach is the difference between the condensing refrigerant temperature and the leaving condenser-water temperature, and as the tubes foul, that gap widens because the heat transfer is getting worse. Trend the approach over weeks and a rising number points straight at fouling or non-condensables, often before the chiller ever surges or its efficiency log raises a flag. The evaporator side has its own approach that tells you about the chilled-water barrel the same way.

The fix is mechanical cleaning. Brushing the tubes, or eddy-current inspection followed by cleaning, restores the heat transfer and drops the condensing pressure and the lift back to where the machine has margin again. Good condenser-water treatment and a clean tower are what keep the approach from climbing in the first place, which is why the tower-cleaning and water-treatment work pays for itself on the chiller side. Hedge the actual approach and pressure targets to the manufacturer's data and the refrigerant, because the right numbers vary by machine.

Low refrigerant charge

A low refrigerant charge can push a machine toward surge too, and it is worth a quick check when the usual high-lift causes come up clean. Short on refrigerant, the evaporator cannot keep the suction pressure up, the effective lift the compressor sees widens, and the machine loses surge margin at the same load it used to hold. Low charge usually shows up alongside other signs, a low suction pressure, a low liquid level in the sight glass, or a chiller that cannot make its leaving chilled-water setpoint, so read it in context rather than in isolation.

Confirm it the right way before adding refrigerant. On a centrifugal that means checking levels and pressures against the manufacturer's procedure, not topping off on a hunch, because overcharging brings its own problems and a leak that took the charge down will take the next charge down too. Find the leak, fix it, then charge to the machine's specification.

How do you diagnose a surging chiller?

Confirm the surge first, then go straight to the lift and the load. The confirmation is the easy part: the cyclic bang, the amp swing in step with it, the discharge pressure and temperature jumping on the same beat. Pull the surge count and the trends off the panel if the machine logs them, because the history tells you whether this is a one-time hard start or a pattern that repeats every time the machine unloads.

Once it is confirmed as surge and not just a noise, check the lift. Look at the condensing pressure and temperature against what the load and the entering condenser-water temperature say they should be. Trend the condenser approach for fouling. Check the purge run-time and look for non-condensables on a low-pressure machine. Confirm the condenser-water flow is up and the tower is delivering water at the temperature it should. Each of these is a way lift gets added, and the one that moved is usually the cause.

Then check the load and the flow path. Was the machine at very low load when it surged, with a single large chiller carrying a light building? Is the evaporator flow up? Is the IGV calibrated and reading true position, and on a VFD machine is the drive responding? The pattern of when it surges narrows it fast: surge at high load on a hot day with a wide approach points at the condenser and the lift, while surge only at the bottom of the load range points at turndown, flow, and the capacity control. Find which side moved, fix that, and stop chasing the other.

How do you stop a chiller from surging?

Bring the lift down and keep the flow up, in that order, and the surge goes away because the operating point comes back inside the envelope. The specific moves follow the cause the diagnosis found, and most of them are maintenance, not capital.

On the lift side: clean the condenser tubes to restore heat transfer and drop the condensing pressure, purge the non-condensables on a low-pressure machine or properly evacuate after any service, fix the cooling tower and the condenser-water flow so the water comes back cold and moves at design rate, and confirm the chilled-water setpoint is not being pushed lower than the job needs. Every one of these lowers the head the compressor has to make.

On the flow and control side: verify the IGV calibration and actual vane position against the command, confirm the VFD is responding and the speed-plus-vane logic is doing its job, and use hot-gas bypass to hold flow at the very bottom of the load range where the other controls run out. If a single large machine keeps surging on light load, the real fix is often staging, running a smaller chiller or a different machine that matches the load instead of forcing one big unit down past its stable turndown. And the rule that overrides all of it: do not keep restarting a machine that trips on repeated surge to push through. Find the cause first.

Proving no surge across the load range at commissioning

Commissioning is where you prove the machine stays off the surge line across the full load range, not just at the design point. A chiller can pass a full-load capacity test and still surge every night at low load, so the surge check has to walk the machine down through part load and confirm the IGV, the VFD, and the staging hold it stable the whole way. The startup-and-commissioning guide treats this as one of the gates the plant has to clear before acceptance.

The factory-authorized technician sets and verifies the surge control during startup, because the surge map and the protection setpoints are the manufacturer's and the warranty rides on them. Your job on the commissioning side is to confirm the machine actually holds across the load profile the building will see, including the light-load and low-lift corners where surge hides, and to document that it did. A surge margin that was never tested below 50 percent load is a surge margin you are taking on faith.

Verify the inputs while you are at it. Confirm the condenser approach is where it should be at startup so you have a clean baseline to trend against later, check that the purge is healthy on a low-pressure machine, and record the surge count as zero on a clean commissioning run. Those baselines are what let an operator three years out tell whether a new surge is a control problem or a slow fouling problem.

What repeated surge does to the machine

Repeated surge kills a centrifugal chiller, and it kills it from the thrust bearing out. Each reversal slams the rotor axially, and the thrust bearing is the part that takes that slam. Ride a machine through surge long enough and the bearing wears, the babbitt smears, end-play opens up, and eventually the rotating assembly is no longer held where the tight internal clearances need it to be. From there the impeller can rub or contact the housing, and a rub at full speed is a rebuild or a scrapped machine.

This is why the surge count and the surge trip are protection, not nuisance. A machine designed for decades of service can be ruined in a single shift of unattended surging, and the repair cost dwarfs the cost of the cleaning or the control fix that would have prevented it.

Treat any surge that repeats as a stop-and-find-the-cause event. One surge on a hard start, logged and gone, is the machine doing its job at the edge. Surge every time it stages down, or surge that the operator clears and restarts through, is the machine being destroyed slowly while someone watches the alarm reset.

Surge in data-center and critical-cooling plants

Critical-cooling plants have their own surge exposure, and it is not where people expect. A data center runs a high, steady load most of the time, which is good for surge margin because the machines stay well-loaded and well to the right of the surge line. The risk lives in the off-design moments: a redundant chiller that has to carry a light load during a failover, a plant in deep turndown on a cold night with free cooling carrying part of the job, or a machine staged in and out as the IT load shifts.

Because the cooling cannot drop, the staging and the surge control have to be proven through every transition, not just at steady state. A surge trip on a machine that was supposed to be the backup is a redundancy that was not really there. VFD machines and good staging logic earn their place here, keeping each chiller in its stable envelope through the load swings that critical cooling creates, and the commissioning has to test those swings rather than assume them.

What to document

Surge is a trend problem as much as an event, so the record is what lets the next person see it coming. Capture the surge events and counts with their dates, the operating conditions at each one, and the lift-side baselines, the condenser and evaporator approach, the purge run-time, the condenser-water entering and leaving temperatures and flow, so a slow climb is visible long before it becomes a bang.

The table below is the short version of the causes, what each one does to the machine, and the fix that addresses it. It doubles as a triage sheet when a machine starts surging and nobody is sure where to look first.

Common mistakes

• Running a machine through repeated surge instead of stopping to find the cause, which wears the thrust bearing and risks the impeller.
• Dirty condenser tubes or non-condensables raising the head, with no approach trend kept to catch the climb.
• A cooling-tower or condenser-water-flow problem driving high lift while the compressor gets blamed.
• Mis-set or out-of-calibration IGV or VFD control letting the operating point cross the surge line.
• No hot-gas bypass or staging plan for very low load, so one large machine surges every light-load night.
• Blaming cold tower water for surge and backing off condenser-water reset, when high lift was the real driver.
• Resetting and restarting through a surge trip, or moving the surge-control setpoints, instead of fixing what moved the operating point.

Field checklist

Standards and references

The surge map and the surge-control logic belong to the chiller manufacturer, and they are the first reference for any surge question. The map, the protection setpoints, the minimum entering-condenser-water temperature, the condenser approach targets, and the charging and purge procedures are all machine-specific, so hedge the pressures and the approach numbers to the manufacturer's data and the refrigerant rather than to any rule of thumb. The factory-authorized technician sets the surge control at startup, and the warranty rides on it.

AHRI Standard 550/590 is the rating standard for water-chilling packages in the United States, defining how capacity in tons and efficiency in kW per ton and IPLV are measured at standard conditions, which is the basis for the part-load behavior where surge margin matters. ASHRAE handles the design side, with ASHRAE 90.1 on energy and the ASHRAE TC 9.9 thermal guidelines that shape data-center plant operation and the load profiles a chiller has to ride without surging. The refrigerant and its safety classification, and the purge requirements on a low-pressure machine, fall under the refrigerant management and machine-room safety rules.

Three things hold across all of it, and they are worth stating plainly: high lift is what causes surge, so find and lower the lift; clean the condenser tubes and purge the non-condensables that raise the head; and never run a chiller through repeated surge, because the thrust bearing and the impeller pay for it. Confirm the specific numbers against the adopted standards edition and the manufacturer's documentation before acting on them.

Units and terms

Surge work crosses a few terms and unit systems, and the same idea reads differently across a manufacturer's manual, a controls screen, and a commissioning report. The terms below are the ones that come up most when a centrifugal is surging.

Surge

A momentary reversal of refrigerant flow through the compressor when the lift exceeds what the impeller can make at that flow, repeating until conditions change

Lift (head)

The pressure difference the compressor must develop between evaporator and condenser; high lift is the primary surge driver

Surge line

The low-flow, high-head boundary on the compressor map; cross it and the machine surges

Choke / stonewall

The high-flow limit on the map where gas reaches sonic velocity and flow cannot increase; rarely reached in chiller service

IGV

Inlet guide vanes, a ring of adjustable vanes at the impeller eye that modulate capacity by pre-swirling the entering refrigerant

Approach

The difference between condensing refrigerant temperature and leaving condenser-water temperature; a rising condenser approach points to fouling or non-condensables

Non-condensables

Air and moisture in the refrigerant circuit that blanket condenser tubes and raise condensing pressure and lift

Hot-gas bypass

A valve that routes hot discharge gas back to the low side to fake a load and hold flow off the surge line at very low load, at an energy penalty

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What are the types of chillers?
• What is a factory authorized chiller startup?
• What does a compressor do in an AC or refrigeration system?
• What is building pressurization?
• What is a change order?